Interactive Visualization of Multimodal Brain Connectivity: Applications in Clinical and Cognitive Neuroscience

Magnetic resonance imaging (MRI) has become a readily available prognostic and diagnostic method, providing invaluable information for the clinical treatment of neurological diseases. Multimodal neuroimaging allows integration of complementary data from various aspects such as functional and anatomical properties; thus, it has the potential to overcome the limitations of each individual modality. Specifically, functional and diffusion MRI are two non-invasive neuroimaging techniques customized to capture brain activity and microstructural properties, respectively. Data from these two modalities is inherently complex, and interactive visualization can assist with data comprehension. The current thesis presents the design, development, and validation of visualization and computation approaches that address the need for integration of brain connectivity from functional and structural domains. Two contexts were considered to develop these approaches: neuroscience exploration and minimally invasive neurosurgical planning. The goal was to provide novel visualization algorithms and gain new insights into big and complex data (e.g., brain networks) by visual analytics. This goal was achieved through three steps: 3D Graphical Collision Detection: One of the primary challenges was the timely rendering of grey matter (GM) regions and white matter (WM) fibers based on their 3D spatial maps. This challenge necessitated pre-scanning those objects to generate a memory array containing their intersections with memory units. This process helped faster retrieval of GM and WM virtual models during the user interactions. Neuroscience Enquiry (MultiXplore): A software interface was developed to display and react to user inputs by means of a connectivity matrix. This matrix displays connectivity information and is capable to accept selections from users and display the relevant ones in 3D anatomical view (with associated anatomical elements). In addition, this package can load multiple matrices from dynamic connectivity methods and annotate brain fibers. Neurosurgical Planning (NeuroPathPlan): A computational method was provided to map the network measures to GM and WM; thus, subject-specific eloquence metric can be derived from related resting state networks and used in objective assessment of cortical and subcortical tissue. This metric was later compared to apriori knowledge based decisions from neurosurgeons. Preliminary results show that eloquence metric has significant similarities with expert decisions